Pseudo-bulk samples from external single-cell dataset

To project external single-cell data onto the DC atlas, first single cells were aggregated to build pseudo-bulk samples to mitigate library size differences. For each cluster defined by the author of single-cell data, the cells were randomly separated into subgroups of a size 15, using Sincast (25). Thus, every pseudo-bulk sample was aggregated from 14 cells and the expression values of each aggregated group were a summation of the subgroup’s expression values for each gene.